trumancai/Revela-code-500M

Model Summary

Revela-code-500M brings the Revela self-supervised code-retrieval objective to the 500 M-parameter Qwen 2.5-0.5B backbone—offering a strong quality/latency middle-ground.
Training used 358 K batches of Stack Overflow, tutorial, and API-doc text.

• Repository: TRUMANCFY/Revela
• Training Dataset: trumancai/revela_code_training_corpus

Other Links

Binary	Description
trumancai/Revela-code-1b	1 B-parameter code-retriever.
trumancai/Revela-code-500M	← current repo
trumancai/Revela-code-135M	135 M-parameter code-retriever.
trumancai/Revela-500M	500 M-parameter Wikipedia retriever.
trumancai/revela_code_training_corpus	Code training corpus.
trumancai/revela_training_corpus	Wikipedia training corpus.

Usage

from mteb.model_meta import ModelMeta
from mteb.models.repllama_models import RepLLaMAWrapper, _loader
import mteb, torch

revela_code_qwen_500m = ModelMeta(
    loader=_loader(
        RepLLaMAWrapper,
        base_model_name_or_path="Qwen/Qwen2.5-0.5B",
        peft_model_name_or_path="trumancai/Revela-code-500M",
        device_map="auto",
        torch_dtype=torch.bfloat16,
    ),
    name="trumancai/Revela-code-500M",
    languages=["eng_Latn"],
    open_source=True,
    revision="a7db96558e7d0cf11a490ec3a48f966196cb0312",
    release_date="2025-04-13",
)
model = revela_code_qwen_500m.loader()

mteb.MTEB(tasks=["AppsRetrieval"])
    .run(model=model, output_folder="results/Revela-code-500M")

License

Citation